Neural network-based optimization of fibres for seismic retrofitting applications of UHPFRC

Abstract

Ultra-high-performance fibre reinforced concrete (UHPFRC) is an advanced construction material that provides new opportunities in the future of the construction industry around the world. Among those new applications, rehabilitation, and seismic retrofitting of existing damaged or non-ductile concrete structures can be highlighted. The main objective of this paper is to optimise the hybrid blend of fibres that allows a previously optimised eco-friendly ultra-high-performance cementitious paste to achieve the ductility requirements for seismic retrofitting applications at lower costs. To meet this goal, two artificial neural network models (ANNs) were created to predict the energy absorption capacity (g) and maximum post-cracking strain (εpc). A total of 50 own experimental campaign data added to 550 published works throughout the world data were used for training purposes by using the R-code language. Once the models were trained and validated, a multi-objective optimisation was used to select the combination of fibres that achieved the limit values of g ≥ 50 kJ/m3 and εpc ≥ 0.3% considering cost constraints. The experimentally validated results indicated that the adequate blend of high strength steel micro-fibres and hooked end normal strength steel fibres fulfil the threshold values at a lower cost.

Keywords:

• UHPFRC
• artificial neural networks
• ductility
• multi-objective optimisation
• energy absorption capacity
• seismic retrofitting applications

Disclosure statement

No potential conflict of interest was reported by the authors.